March 25, 2004
Baba vs. Dada
Back in the days when I taught a Phonetics class (because I was in a department that had no genuine phonetician, the kind of person who is not a technophobe and can introduce students to the wonders of phonetics software), I used to give my students an emphatic warning: when you work on your term project, I told them, do tape-record your consultant pronouncing a 200-word Swadesh list of basic vocabulary, but don't use those tapes as a substitute for face-to-face elicitation and checking of data. The reason is that seeing your consultant pronounce the sounds helps you hear them better and identify them correctly. Yesterday I began to doubt the complete wisdom of this advice when my colleague Pam Beddor showed a video in which a lecturer illustrated the McGurk effect. Probably all my fellow bloggers already know about this remarkable demonstration, but I'll describe it anyway.
The speaker announced that she would pronounce a nonsense word, baba. She instructed her audience to close their eyes and listen. Sure enough, with your eyes closed, you could tell that she was saying baba. No surprise there. Then the audience was told to listen again with open eyes. This time the video showed the speaker apparently pronouncing dada -- no lip closure at all, though I couldn't actually see much of what was going on behind the teeth. And in fact I heard dada. No matter how hard I tried, knowing that she was actually saying baba, I could not hear baba. True, it sounded like a slightly odd version of dada, or at least I imagined that it sounded oddish, but I couldn't even imagine baba while watching her. Moral (?): in a clash between eyes and ears, the eyes have it.
[Update by Mark Liberman: Sally is right to be impressed by the McGurk effect -- it's a stunning demonstration of the power of "sensory fusion" in speech perception. However, her description of the details is a bit different from the way in which the standard effect is usually demonstrated. The standard McGurk effect involves seeing a video of [ga] while listening to a synchonized audio of [ba] and perceiving [da], unless you close your eyes. It feels like you're controlling the playback with your eyelids.
There's a excellent McGurk page
here.]
